Inference Related to Common Breaks in a Multivariate System with Joined Segmented Trends with Applications to Global and Hemispheric Temperatures

Dukpa Kim, Tatsushi Oka, Francisco Estrada, Pierre Perrón

Abstract

Recent research suggests that global temperatures and radiative forcing are characterized by linear trends with two distinct changes in growth rates. The first, occurring in the early 1960s, marks the "onset of sustained global warming." The second, occurring in the mid-1990s, relates to the so-called "hiatus period," where the rate of increase slowed. This paper addresses two critical unresolved issues: whether these breaks in slope are common across temperature and radiative forcing series, and whether the mid-90s hiatus can be formally established as a structural shift rather than a product of natural internal variability. By developing new multivariate statistical tests for joined segmented trends, the authors demonstrate that these breaks are indeed common and statistically significant, providing strong evidence for a causal link between anthropogenic factors and global temperature shifts.

Introduction

Climate science has long grappled with the non-linear nature of global warming. Rather than a steady climb, temperature records exhibit "segments" with varying growth rates. Two periods have generated significant debate: the sharp acceleration of warming in the 1960s and the apparent slowdown, or "hiatus," beginning in the 1990s.

Proving that these shifts are directly caused by human activity requires more than just observing parallel lines on a graph. In time-series analysis, variables can appear related by chance—a phenomenon known as spurious correlation. This paper provides a rigorous statistical framework to determine if the dates of these trend changes (the "breaks") are identical across both temperature and anthropogenic radiative forcing. If the "thermostat" (forcing) and the "room" (temperature) shift at the exact same moment, the case for causality becomes scientifically formidable.

Methodological Framework

The authors treat the climate as a multivariate system where temperatures and radiative forcing are interdependent variables. They introduce several innovations to traditional time-series analysis:

Unlike standard models that allow for abrupt "jumps" in data, this study utilizes a model where the trend is continuous (joined) at the break date. This approach is more physically realistic for climate systems, where temperatures do not instantaneously leap but rather change their rate of accumulation.

New Multivariate Tests

The paper introduces three primary tests to detect common breaks across multiple series:

  • Quasi-Likelihood Ratio (LR) Test: A robust standard approach for identifying structural shifts.
  • GLS-Wald and OLS-Wald Tests: These are corrected versions of the Wald test. The OLS-Wald offers computational efficiency, while the GLS-Wald accounts for the correlation between different equations—such as the synchronized fluctuations between the Northern and Southern hemispheres.

Addressing Noise and Variability

To ensure the tests do not "hallucinate" breaks due to natural noise, the authors employ a Bootstrap Procedure to generate reliable p-values. Furthermore, they use the Atlantic Multidecadal Oscillation (AMO) and North Atlantic Oscillation (NAO) indices to "purge" the temperature data of low-frequency natural variability. This filtering allows the underlying secular trend driven by human activity to emerge more clearly.

Data and Analysis

The study utilizes a comprehensive suite of datasets to ensure robustness:

  • Temperature Series: HadCRUT4, GISS-NASA, Berkeley Earth, and the Karl et al. (2015) datasets, covering global and hemispheric data from as early as 1850.
  • Radiative Forcing: GISS-NASA data focusing on Well-Mixed Greenhouse Gases (WMGHG) and Total Radiative Forcing (TRF).
  • Natural Oscillations: AMO and NAO indices used for noise filtering.

Key Findings

1. The 1960s Onset

The tests confirm a common break in the early 1960s (specifically between 1963 and 1966). This period marks the definitive onset of sustained global warming, directly corresponding to a surge in greenhouse gas forcing.

2. The Reality of the 1990s "Hiatus"

One of the paper's most significant contributions is the formal rejection of the idea that the post-1990s warming slowdown was merely natural variability. The authors identify a common break around 1990–1992 where the rate of growth in both forcing and temperature significantly decreased. By testing temperature and forcing together (a bivariate approach), the authors achieved higher statistical power than previous studies, allowing them to detect this slowdown even in noisy datasets.

3. Commonality as Evidence of Causality

The results show that for filtered series, the null hypothesis of "no common break" was rejected across nearly all pairs of temperature and forcing data. This synchronization suggests that the shifts in temperature trends are not internal fluctuations but are driven by the underlying shifts in radiative forcing.

Conclusion

By providing a formal testing procedure that accounts for noise and joined trends, this paper settles a major point of contention in climate science. The findings confirm that the major turning points in 20th and 21st-century climate history are shared between human-induced forcing and global temperatures. Beyond climate science, the multivariate joined-trend tests developed here offer a valuable tool for economics and other fields where shared structural shifts are suspected across multiple variables.

Summary updated

Abstract

What transpires from recent research is that temperatures and radiative forcing seem to be characterized by a linear trend with two changes in the rate of growth. The first occurs in the early 60s and indicates a very large increase in the rate of growth of both temperature and radiative forcing series. This was termed as the "onset of sustained global warming". The second is related to the more recent so-called hiatus period, which suggests that temperatures and total radiative forcing have increased less rapidly since the mid-90s compared to the larger rate of increase from 1960 to 1990. There are two issues that remain unresolved. The first is whether the breaks in the slope of the trend functions of temperatures and radiative forcing are common. This is important because common breaks coupled with the basic science of climate change would strongly suggest a causal effect from anthropogenic factors to temperatures. The second issue relates to establishing formally via a proper testing procedure that takes into account the noise in the series, whether there was indeed a `hiatus period' for temperatures since the mid 90s. This is important because such a test would counter the widely held view that the hiatus is the product of natural internal variability. Our paper provides tests related to both issues. The results show that the breaks in temperatures and radiative forcing are common and that the hiatus is characterized by a significant decrease in their rate of growth. The statistical results are of independent interest and applicable more generally.
BibTeX
@article{kim2019inference,
  author = {Dukpa Kim and Tatsushi Oka and Francisco Estrada and Pierre Perrón},
  title = {{Inference Related to Common Breaks in a Multivariate System with Joined Segmented Trends with Applications to Global and Hemispheric Temperatures}},
  journal = {Journal of Econometrics},
  year = {2019},
  doi = {10.1016/j.jeconom.2019.05.008},
}
Figure